24/192 Music Downloads and why they make no sense - Page 4 - AVS Forum
Forum Jump: 
Reply
 
Thread Tools
post #91 of 761 Old 03-10-2012, 07:54 PM
AVS Special Member
 
hd_newbie's Avatar
 
Join Date: Jun 2008
Location: Metro DC
Posts: 2,088
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Quote:
Originally Posted by amirm View Post

So you are saying Meyer and Moran lied when they said this: ""...virtually all of the SACD and DVD-A recordings sounded better than most CDs sometimes much better. Had we not degraded the sound to CD quality and blind-tested for audible differences, we would have been tempted to ascribe this sonic superiority to the recording processes used to make them."

Come on. Say it. Do you think they were wrong when they said the formats sounded better to them?

That part made no sense to me and still doesn't.

By "other things equal", I was trying to say, one can not measure the impact of media quality when there are other variables such as recording quality and different kinds of mastering. Arny explained it for me.

I do not deny SACDs are typically better mastered, but one can reach identical quality with a CD if desires so.

For the rest, let's agree to disagree, shall we?
hd_newbie is offline  
Sponsored Links
Advertisement
 
post #92 of 761 Old 03-11-2012, 04:42 AM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by Ethan Winer View Post

Dude, you need better software. I just did the same in Sound Forge and got much less distortion. I created a 100 Hz tone at full scale with 24 bits, then truncated that to 16 bits with no dither. In the truncated version none of the artifacts are louder than -120 dB full scale.

Actually, Amir made yet another rookie mistake. Here's the graph that he clams is the result of trunctation to 16 bits:



The tip off is that the spikes peak into the -45 dB region. I can duplicate Amir's results by truncating to 8 bits. This is actually a pretty gross error on his part - characterizing 8 bit truncation as being 16 bit truncation. It's typical of Amir to make 16 bits look far worse than they really are. In this case, I'm going to be generous and characterize them as being an honest mistake that his biases and inexperience predisposed him to.
arnyk is offline  
post #93 of 761 Old 03-11-2012, 09:08 AM
Senior Member
 
rock_bottom's Avatar
 
Join Date: Apr 2004
Posts: 430
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
I think I see the reason for the distortion discrepancy. Amir's post pointed back to Vincent Kars' site, which in turn links to this PDF file. Said PDF file is the origin of the two plots. Looking at the text of the file, it says this:

Quote:
Originally Posted by Nika Aldrich View Post

The first plot is the original 24 bit waveform on a spectragraph (amplitude on the Y axis, frequency on the X axis). The second plot is the truncated version of this. You'll see in the truncated waveform a series of fairly high peaks at around 36dB down from the signal's peak. This is because this was a low level waveform to start with (the signal was recorded low to make the explanation more visibly obvious). If this were a "fully maximized" signal, (one that was recorded at full scale), these peaks would be much lower.

The author says it is a "low level waveform", but unfortunately does not specify the level. So apparently, the dB scale is not dBFS.

The distortion (due to the quantization noise being correlated with the signal) gets worse as the signal level decreases because the amplitude excursion of said quantization noise is fixed at +/- 1/2 LSB. So when you decrease signal amplitude, the quantization noise amplitude becomes a larger percentage of the signal amplitude.
rock_bottom is offline  
post #94 of 761 Old 03-11-2012, 10:41 AM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by rock_bottom View Post

I think I see the reason for the distortion discrepancy. Amir's post pointed back to Vincent Kars' site, which in turn links to this PDF file. Said PDF file is the origin of the two plots. Looking at the text of the file, it says this:

The author says it is a "low level waveform", but unfortunately does not specify the level. So apparently, the dB scale is not dBFS.

The chart is clearly made by Spectrafoo, which is a well-known and well-documented device. From the perspective of Spectrafoo, the 0 dB indication at the top left of the display represents 0 dB FS.


Quote:


The distortion (due to the quantization noise being correlated with the signal) gets worse as the signal level decreases because the amplitude excursion of said quantization noise is fixed at +/- 1/2 LSB. So when you decrease signal amplitude, the quantization noise amplitude becomes a larger percentage of the signal amplitude.

If you take this plot at face value, the test signal was -6 dB FS, which is a common reference level for experiments like this. Whether this signal is a "low level signal" or a "medium level signal" could be debated. There would be no significant cost in clarity if the signal were at a lower amplitude, and shown with a corrrect postition with respect to FS.

If someone knows what they are doing they can redo this experiment with their own software or hardware, and produce plots that remove all possible ambiguity.

Amir's shifting of the discussion from reconstruction filtering to dither was completely off topic. Picked an unfortunate reference is again characteristic of people with agendas and inexperience. Adding dither does not obscure of conceal the purported stepwise nature of digital signals because the staircase just moves the jitter signal around. The usual kinds of low pass filtering that is done does obscure or conceal the steps quite effectively.
arnyk is offline  
post #95 of 761 Old 03-11-2012, 11:23 AM
Senior Member
 
rock_bottom's Avatar
 
Join Date: Apr 2004
Posts: 430
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 14
The author really should have mentioned what the signal level was, so that people could duplicate the experiment properly if they wanted.

But yes, using the bad results of an undithered conversion as justification for the alleged audible superiority of high-res is a red herring. One assumes that the conversion is competently performed.
rock_bottom is offline  
post #96 of 761 Old 03-11-2012, 11:29 AM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by arnyk View Post

Amir's shifting of the discussion from reconstruction filtering to dither was completely off topic. Picked an unfortunate reference is again characteristic of people with agendas and inexperience. Adding dither does not obscure of conceal the purported stepwise nature of digital signals because the staircase just moves the jitter signal around. The usual kinds of low pass filtering that is done does obscure or conceal the steps quite effectively.

Lots of fancy words indicating you are dabbling in audio alchemy than real science . We can show these effects in pictures much easier than audio since we can freeze a frame and look at it.

I will use the image in the Wiki for dither.http://en.wikipedia.org/wiki/Dither



The first image is the original. The second from left is reducing its color pallet to 216 "web safe" set. We can see the massive distortion we created by reducing our color steps. Now I apply the low pass filter Arny mentions, going even further as to soften the image itself -- something you don't want to do to your audio -- and what do you get? The same distortions albeit, a bit softer but they are distinctly there. Now look at the last image which is with proper application of dither with the same 216 colors. You think filtering was a better solution than dither as Arny says? I think not. Can you see the "steps" (contour edges) any more in the properly dithered version? Nope. Can you still see them in the filtered Arny version? Yes.

Once you butcher a signal, you don't get to make up for it by filtering or we would filter everything to remove distortion. You think if I put a 22 Khz low pass filter on my amp output, I get rid of its clipping distortions? If I could, that would be a standard thing on every speaker and you would not have to worry about your amp ever clipping! Sadly, audio alchemy doesn't work.

This back and forth demonstrates the level of confusion that exists here. If someone like Arny thinks that he doesn't need to dither because the reconstruction filter in every DAC takes care of the distortion when in reality does no such thing, we have a potential problem with people doing conversions from professional recordings to 16 bits. It is best to get Arny's of the world out of the loop and give the consumer the original bits and not assume that every music producer understands digital signal processing.

I had a similar discussion with a Blu-ray Disc producer here a few years back who thought dither=noise and said he would have nothing to do with it on principal. The way dither works is not intuitive to folks so we get people making the wrong choices like this. Same happens in video. When film is digitized ("telecine") it is captured at 10 bits whereas DVD and BD are 8 bits. You won't believe how many people just truncate down to 8 bits and cause contouring like the above images. It is not like dither is a new concept as it has been around for decades. Yet major productions of HD content still get it wrong. So I say let's get the person in the middle out of the way. Our video needs to be 10 bits and let's release the darn music in 24 bits and be done with it.

P.S. Thanks Rock for chiming in

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #97 of 761 Old 03-11-2012, 01:08 PM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 133
Quote:
Originally Posted by arnyk View Post

The tip off is that the spikes peak into the -45 dB region. I can duplicate Amir's results by truncating to 8 bits.

Good catch Arny.

RealTraps - The acoustic treatment experts

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Ethan Winer is offline  
post #98 of 761 Old 03-11-2012, 01:10 PM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 133
Quote:
Originally Posted by amirm View Post

We can show these effects in pictures much easier than audio since we can freeze a frame and look at it.

This is great for explaining the theory, but it's useless in practice because truncation distortion at 16 bits is inaudible. Versus color reduction to only 8 bits which obviously is visible. I think this logic error is called a Red Herring.

--Ethan


RealTraps - The acoustic treatment experts

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Ethan Winer is offline  
post #99 of 761 Old 03-11-2012, 01:55 PM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by amirm View Post

Lots of fancy words indicating you are dabbling in audio alchemy than real science . We can show these effects in pictures much easier than audio since we can freeze a frame and look at it.

I will use the image in the Wiki for dither.http://en.wikipedia.org/wiki/Dither



Trying to conflate pictures with sound when it comes to understanding concepts like dithering, quantization and spectral response is yet another well-known rookie mistake.
arnyk is offline  
post #100 of 761 Old 03-11-2012, 01:58 PM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by amirm View Post

Once you butcher a signal, you don't get to make up for it by filtering or we would filter everything to remove distortion.

The rookie mistake here is conflating problems in the amplitude domain with problems in the frequency domain.
arnyk is offline  
post #101 of 761 Old 03-11-2012, 02:45 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by arnyk View Post

The rookie mistake here is conflating problems in the amplitude domain with problems in the frequency domain.

http://en.wikipedia.org/wiki/Fourier_transform

"The Fourier transform is a mathematical operation with many applications in physics and engineering that expresses a mathematical function of time as a function of frequency, known as its frequency spectrum; Fourier's theorem guarantees that this can always be done."

If you don't know that the time domain representation of a signal is inexplicably connected to its frequency domain representation, you are arguing that 1+1 is not equal to 2.

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #102 of 761 Old 03-11-2012, 02:57 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by arnyk View Post

Trying to conflate pictures with sound when it comes to understanding concepts like dithering, quantization and spectral response is yet another well-known rookie mistake.

Not knowing that the mathematics of dither/low-pass filter has no idea that the numbers it is dealing with is audio versus picture pixels, is another .

http://en.wikipedia.org/wiki/Dither

In digital processing and waveform analysis

Dither is often used in digital audio and video processing, where it is applied to bit-depth transitions; it is utilized in many different fields where digital processing and analysis are used especially waveform analysis. These uses include systems using digital signal processing, such as digital audio, digital video, digital photography, seismology, RADAR, weather forecasting systems and many more.

The premise is that quantization and re-quantization of digital data yields error. If that error is repeating and correlated to the signal, the error that results is repeating, cyclical, and mathematically determinable. In some fields, especially where the receptor is sensitive to such artifacts, cyclical errors yield undesirable artifacts. In these fields dither results in less determinable artifacts. The field of audio is a primary example of this the human ear functions much like a Fourier transform, wherein it hears individual frequencies. The ear is therefore very sensitive to distortion, or additional frequency content that "colors" the sound differently, but far less sensitive to random noise at all frequencies."


As you see, the section in blue makes no distinction between data types. Audio is just given as an example as is imaging which is where most of my example came from. If these were entirely different concepts, they would have different Wiki pages.

Of course, when it comes to specific algorithms for dither, we might choose different ones based on perceptual effect but nothing in the theory or the simulation I showed changes in that regard. Nice attempt at FUD though.

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #103 of 761 Old 03-11-2012, 03:10 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by Ethan Winer View Post

This is great for explaining the theory....

Ah wait . The theory is important. Because, the person is not just saying what you are saying. He is going on to say "it is because...." If that part is incorrect, then the person's summation of the issues at hand is not correct. They need to learn more and with it, maybe discover new ways to detect audibility. If you don't know *why* something is the way it is, then you don't understand it well enough. In this case, I am referring to Arny saying quantization noise manifests itself as multiples of sampling frequency and that we can filter it out with the reconstruction filter.

Quote:


... but it's useless in practice because truncation distortion at 16 bits is inaudible. Versus color reduction to only 8 bits which obviously is visible. I think this logic error is called a Red Herring.

I respect your opinion as someone who works in the field and based on experimentation you have conducted Ethan. Problem is, you can't prove it to everyone's satisfaction.

This is why I keep saying that if you/talent get to listen to 24-bit tracks yourself why when it comes time to giving me the music, you think you should stuff it with distortion. It is no longer the issue of what the CD spec requires. Why would you want to be in the position of explaining that stuffing 40+db of distortion is just dandy?

Think of your customers. If a butcher spat on their stakes before he gave it to them because he thinks the germs would burn off when they cook them, would that be OK?

Do you archive your recordings when you are done in 16 bits? I suspect not. I assume you are storing them as 24-bit. If so, why wouldn't a customer want the same bits?

Quote:



.

BTW, the picture on the left is not an engineer but likely a technician. At least not one in the last two decades who spend most of their time designing using a computer and have a tech do board level mods, and outside shop doing the fabrication and assembly.

Reminds me of a story. I was at Sony and I had a group designing VLSI/computers hardware. I hear Chairman Morita-san was coming over to visit us. Prior to him arriving one of his handlers showed up and asked where the engineers were. I point to the cubicals with my crew sitting in front of workstations and said, "there." He could not believe it and kept asking where the "lab" was that would work in. I guess his view of engineers is this distorted one from Verizon:



Thank heavens they didn't have him wear a protractor!

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #104 of 761 Old 03-11-2012, 03:26 PM
AVS Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,908
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by amirm View Post

I do better per above. I don't rely on wikipedia if that is what you mean .

Quote:
Originally Posted by amirm View Post

http://en.wikipedia.org/wiki/Dither

Quote:

When someone uses a wiki citation to bust one of your incorrect assertions, wiki is crap. Now you cite it freely? Double-standard much?

Also the cat image with severe contouring is reduced from a 16.7million color palette (8 bits/component, 24bit total) to a 256 color palette. That’s more like <3 bits / RGB component, 8 bit total.

Our video world is full of images with 8 bit per component, (actually more like 7.7bits in-gamut with StudioRGB) and they are for the most part, banding-free to human perceptions. Nothing like the cat image you show that has been severely reduced to a 256 color palate.

I haven’t noticed ANY contouring on the last 20-30 BD’s I’ve watched at 8 bits/component. Examples of human perceptual contouring with 8bit/component real-world consumer images are exceedingly rare.



Quote:
Originally Posted by amirm View Post

BTW, the picture on the left is not an engineer but likely a technician.

How about legendary analog EE's like Bob Pease and Jim Williams.

dlarsen is offline  
post #105 of 761 Old 03-11-2012, 03:27 PM
AVS Addicted Member
 
Ratman's Avatar
 
Join Date: May 2002
Location: Collingswood, N.J.
Posts: 14,455
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 167 Post(s)
Liked: 285
Ratman is offline  
post #106 of 761 Old 03-11-2012, 04:12 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by dlarsen View Post

When someone uses a wiki citation to bust one of your incorrect assertions, wiki is crap. Now you cite it freely? Double-standard much?

If you know enough to know if the Wiki citation is true or not, then it is fine to use it. I raise concern when someone just puts that forward without even knowing what it says let alone if it is correct. If you have doubt regarding the validity of the two pieces I quoted, I welcome your challenge with specifics.

Quote:


Also the cat image with severe contouring is reduced from a 16.7million color palette (8 bits/component, 24bit total) to a 256 color palette. That’s more like <3 bits / RGB component, 8 bit total.

Yet the dithered version looked remarkably good which shows what happens when you understand signal processing and put it to good use. But that wasn't the point. The point was the middle picture where Arny said as a matter of theory signal processing, low pass filtering fixes quantization noise. It did not do that. Despite all the defensiveness you all are showing, I was only covering the theory there, not attempting to quantify it with respect to audibility.

Quote:


Our video world is full of images with 8 bit per component, (actually more like 7.7bits in-gamut with StudioRGB) and they are for the most part, banding-free to human perceptions. Nothing like the cat image you show that has been severely reduced to a 256 color palate.

Be careful that you don't mix color pallet with color resolution. That aside, the simulations there which were from the Wiki page and not me, show what happens if you take your true color 24-bit samples, and convert them to just use the web-safe colors without dither. They will indeed look that bad if you did not apply dither to it.

You are right that you don't see continuous tone images that way on the web because we prefer to use a work around than to deal with that: compressed image formats like JPEG. Alternatively, dither is used to hide said artifacts. Here is the example dialog box in Photoshop for saving for web which I use very often to reduce file sizes:



Note the color pallet and dither control.

Quote:


I haven’t noticed ANY contouring on the last 20-30 BD’s I’ve watched at 8 bits/component. Examples of human perceptual contouring with 8bit/component real-world consumer images are exceedingly rare.

Not all content shows it, and shows it well enough for everyone to notice. It is noted in reviews though: http://www.blu-ray.com/movies/Kickboxer-Blu-ray/4911/

".... the transfer features minor banding...."

http://bluray.highdefdigest.com/5782/tree_life.html

".... slight banding is fairly noticeable on the outskirts of the objects coming into or vanishing from view."

http://www.blu-ray.com/movies/2012-Blu-ray/8339/

"....Aside from slight visible banding in a few select shots...."

Truth to be told, banding is far more frequent than people note them (even reviewers). As you say, most people don't notice them. Otherwise they would be up in arms over them. We used to be in the loop for production of every HD DVD (and many BDs from dual format studios) and we would see them very frequently but even people doing the encoding would not. In one case, the animation of a studio logo had this issue which was the beginning of every disc they made! They dug into it and you guessed it, it was lack of dither in decimation of the samples.

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #107 of 761 Old 03-11-2012, 05:11 PM
AVS Special Member
 
dlarsen's Avatar
 
Join Date: Mar 2001
Location: Beaverton, OR, USA
Posts: 1,908
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 11
Quote:
Originally Posted by amirm View Post

If you know enough to know if the Wiki citation is true or not, then it is fine to use it. I raise concern when someone just puts that forward without even knowing what it says let alone if it is correct.

So any wiki reference that you make is OK, because you know enough. Any wiki reference that someone else makes, that refutes you, is bogus. Because they don't know enough, like you do. I see.

Quote:
Originally Posted by amirm View Post

If you have doubt regarding the validity of the two pieces I quoted, I welcome your challenge with specifics.

I don't doubt the validity of the two pieces you quoted. Just noting your play-rules for wiki usage.
dlarsen is offline  
post #108 of 761 Old 03-11-2012, 05:58 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by dlarsen View Post

So any wiki reference that you make is OK, because you know enough.... don't doubt the validity of the two pieces you quoted. Just noting your play-rules for wiki usage.

Given your confirmation of the correctness of my posting, you still thought it was appropriate to jump in and complain? Isn't this then a good use of quoting Wiki?

Quote:


Any wiki reference that someone else makes, that refutes you, is bogus. Because they don't know enough, like you do. I see.

You don't strike me as the shy type. So how about next time someone posts a Wiki and I complain, come and defend their use of it technically. If you don't, I assume I am OK to complain. If you do jump in, then you and I will discuss it.

Hope this settles it and we can go back to discussing audio rather than me .

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #109 of 761 Old 03-12-2012, 05:34 AM
AVS Addicted Member
 
arnyk's Avatar
 
Join Date: Oct 2002
Location: Grosse Pointe Woods, MI
Posts: 14,381
Mentioned: 2 Post(s)
Tagged: 0 Thread(s)
Quoted: 747 Post(s)
Liked: 1161
Quote:
Originally Posted by amirm View Post

Not knowing that the mathematics of dither/low-pass filter has no idea that the numbers it is dealing with is audio versus picture pixels, is another .

As usual Amir, you're changing the subject in order to avoid admitting your error.

Here's the problem you keep running from, Amir a full quote of what you said:

Quote:
Originally Posted by amir View Post

For others who don't know what we are talking about, when digital systems use fixed bit depth, you create distortion since analog waveforms do not have such steps. The way to solve this is to add some amount of random data in the form of noise to the signal prior to digitalizing it. That converts distortion into noise which is a good thing. What is not a good thing is that we have now raised the noise floor of our system relative to the source. Noise shaping says instead of adding noise that has equal power in the audible frequency band, we make it so that it has the least amount where the ear is most sensitive and more of it where it is not.

The steps are not removed by dither, they are removed by low pass filtering in the reconstruction filter.

Amir, you've compounded your rookie errors by citing a highly flawed source of misinformation about dither. It purported to show the effects of 16 bit sampling, when in fact it showed the effects of 8 bit sampling.

If we can get you to own up to your own mistakes the discussion can proceed.
arnyk is offline  
post #110 of 761 Old 03-12-2012, 08:21 AM
AVS Special Member
 
Swampfox's Avatar
 
Join Date: Mar 2001
Posts: 3,711
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 13
Quote:
Originally Posted by arnyk View Post


The steps are not removed by dither, they are removed by low pass filtering in the reconstruction filter.

Sometimes the two of you argue semantics.

The steps are removed by filtering.
The quantization error is not.
Truncating a digital signal causes truncation error which is quantization error.
Quantization error causes correlated distortion.
Dither essentially decorrelates the quantization distortion.

Whether quantization distortion is audible in a well produced 44.1/16 bit audio tract is debatable.
Swampfox is offline  
post #111 of 761 Old 03-12-2012, 09:07 AM
Senior Member
 
Mr Man's Avatar
 
Join Date: Oct 2007
Posts: 228
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 10
'While the doctors consult, the patient dies'.

Popcorn and dictionary...
Mr Man is offline  
post #112 of 761 Old 03-12-2012, 09:45 AM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by arnyk View Post

As usual Amir, you're changing the subject in order to avoid admitting your error.

There was no error. I explained what happens when you capture something. You added the bit on how it plays and then proceeded to call it incorrect.

Quote:


The steps are not removed by dither, they are removed by low pass filtering in the reconstruction filter.

They are not removed by low pass filter. I demonstrated that with pictures. Since you don't like that, please show how they are removed in the author's real measurements:



If the steps are removed, why are those distortion spikes still there? Isn't it because the reconstruction filter for 44.1 Khz sampling is at 22.05 Khz and therefore it cannot possibly remove distortions at 300, 500, 700 Hz and such as shown in the above measurement?

Quote:


Amir, you've compounded your rookie errors by citing a highly flawed source of misinformation about dither. It purported to show the effects of 16 bit sampling, when in fact it showed the effects of 8 bit sampling.

Not at all. You didn't appreciate the points I and Rock_bottom made. You proceeded to add your own information to his measurements such as it being a -6 db down from full scale digital whereas he clearly said it was a "low level signal" A -6dbfs is essentially a full scale signal. There is nothing about that is "low level."

If we follow the test conditions as he describes, then we see that his measurements are real. Let's first prove that what we are looking at is NOT a simulation in software but real *analog* measurements of the output as I mentioned.

First his measurements:



Now let's compare it to the software simulation of the same by Ethan with a -9dbfs signal:



We immediately see lots of differences:

1. The noise floor is not flat and is some 40 db higher than Ethan's simulations. We know that real DACs cannot achieve 24 bits of dynamic range and that is the case with author's measurement.

2. You see those two little peaks after the original tone? They are at 200 Hz and 300 HZ with the original tone being at 100Hz. What are they? Harmonic distortions created by a *real* DAC. They do not exist in Ethan's simulations because he assumes an ideal DAC.

So we know that what he did was that he played a low level signal and then made a measurement on the *analog output of his system.* This is important because we very well could have had a gain stage in either the playback system or the measurement.

As Rock mentioned, the level of quantization distortion remains the same when you decimate a signal down to 16 bits but relative to a low level signal, its strength can sharply increase as a percentage of it. So I simulated what would happen if you took a ~-50db fs "low level" signal, in my case at 32 bit samples, and chopped it down to 16 bits (NOT 8). Then I amplified it so that the amplitude of the main 100 Hz signal was similar to the author's. This is what I got:



Let's compare it now to author's real life measurements:



While I did not try to match my simulations 100% to his, and his is a real device and mine is a simulation, you can see that in both cases even though our samples are 16 bits, our distortion relative to our signal has risen way higher. It is a simple matter of what happens when our source signal is low in level and you amplify it and the noise with it as you play.

The situation demonstrates what I always say about digital: it is a superb system for reproducing loud signals. If you want to see what a digital system doesn't do as well, you need to look at low level signals. In this scenario, if someone recorded music at -50db fs and you turned up the volume to listen to it, you would be hearing the distortion which in this case would serve to make the sound harsher as it adds all of those higher frequency harmonics -- kind of like what happens if your amplifier was clipping.

Remember again, I was not trying to demonstrate anything here but the fact that dither does what it says it does. That was author's goal and was also mine. That the dithered signal had none of these spikes.

Quote:


If we can get you to own up to your own mistakes the discussion can proceed.

You not only think I am wrong, but a third-party that did the above tests and another person who quoted it. Earlier it was Bob Stuart that was also wrong, writing corporate white paper rather than something published by AES. So it is not just me that you think is wrong but others not in the conversation.

On the technical front, you imply that the DAC low pass filtering which exists *above* our hearing range gets rid of harmonics *in* our hearing range. This is an impossibility of course unless you are playing a semantic game of saying they are not pure square waves. Yous said that distortions in time domain do not map to the same in frequency domain, violating the most fundamental theories of signal processing (Fourier). It is important that we get the theory right here as that is *objective*.

Earlier in the thread, you had no trouble believing Monty's statement that ultrasonic tones create audible results in the audible frequency band proven by some listening test. I questioned it since it had no backing. I explained how a full amplitude 26 Khz signal was not a realistic situation. So in both cases, I was taking the position of inaudibility but you were the believer. I could have really ran with his statement there. But it was not right to do that based on what he was saying.

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #113 of 761 Old 03-12-2012, 10:10 AM
AVS Special Member
 
Ethan Winer's Avatar
 
Join Date: Apr 2003
Location: New Milford, CT, USA
Posts: 5,748
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 7 Post(s)
Liked: 133
Quote:
Originally Posted by amirm View Post

I respect your opinion as someone who works in the field and based on experimentation you have conducted Ethan. Problem is, you can't prove it to everyone's satisfaction.

That's certainly true. But the problem is them, not me. Some people will not be convinced no matter how compelling and logical the argument. Remember that pathetic exchange at your forum between me and Mike Lavigne when he claimed that digital misses capturing the "essence" of music that vinyl does capture? His logic and understanding of audio is equal to the unicorns and fairies in my drawing. No matter what I explained, in great detail, he could not accept it because of his ignorance and prejudice. I'm sure even you understand that digital is far more accurate than vinyl, and misses nothing, even if some people prefer the sound of vinyl. Yet I'm the one you banned rather than him and those who think and throw insults like him.

Quote:


if you/talent get to listen to 24-bit tracks yourself why when it comes time to giving me the music, you think you should stuff it with distortion.

That's a poor analogy, as is the comparison to a chef spitting on food, but I'll humor you anyway. If the distortion is 90+ dB below the music, and saves half the bandwidth and disk space, where's the damage? You do understand that distortion at -90 dB is around 0.003 percent, yes? What speakers do you know of that won't drown that out by two orders of magnitude?

Quote:


BTW, the picture on the left is not an engineer but likely a technician.

I spent half an hour trying to find a photo of a nerd in a lab coat with a pocket protector and slide rule. Alas, the photo I used was the closest I could find. But it makes the point well. Many 'phooles really do believe that audio gear is designed by fairies and unicorns.

--Ethan

RealTraps - The acoustic treatment experts

To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

Ethan Winer is offline  
post #114 of 761 Old 03-12-2012, 11:04 AM
AVS Special Member
 
hd_newbie's Avatar
 
Join Date: Jun 2008
Location: Metro DC
Posts: 2,088
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Quote:
Originally Posted by Swampfox View Post

Whether quantization distortion is audible in a well produced 44.1/16 bit audio tract is debatable.

It is not audible until someone proves that it is. Without any evidence, all this is just audiophile boo boo.
hd_newbie is offline  
post #115 of 761 Old 03-12-2012, 11:07 AM
AVS Special Member
 
hd_newbie's Avatar
 
Join Date: Jun 2008
Location: Metro DC
Posts: 2,088
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 16
Quote:
Originally Posted by Ethan Winer View Post

Remember that pathetic exchange at your forum between me and Mike Lavigne when he claimed that digital misses capturing the "essence" of music that vinyl does capture? His logic and understanding of audio is equal to the unicorns and fairies in my drawing.

I am not surprised. This is the same guy who accepted a challenge that he could tell different cables apart blindly. He argued he passed all the tests he conducted at home and invited others to prove his case. Do you want to take a wild guess about what happened when Chris went to his home and conducted the test blindly?

After he made a statement saying he would re-visit his beliefs in audio, but some things never change.

Quote:
Originally Posted by Ethan Winer View Post

I'm sure even you understand that digital is far more accurate than vinyl, and misses nothing, even if some people prefer the sound of vinyl.

Does he really? I like to hear that from Amir. I wouldn't be too sure if I were you.

Quote:
Originally Posted by Ethan Winer View Post

Yet I'm the one you banned rather than him and those who think and throw insults like him.

hmmm? How come I thought "WBF" from was very civil? We have been getting daily lectures and advertisements about how civil and great a place it is. You must be making this up!

That being said, of course it will be you who is banned. Think about it who is more likely to buy stuff from Madrona Digital? You or folks like Mike Lavigne. I am not saying this to be despiteful btw. It is just business.
hd_newbie is offline  
post #116 of 761 Old 03-12-2012, 11:46 AM
AVS Special Member
 
Swampfox's Avatar
 
Join Date: Mar 2001
Posts: 3,711
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 0 Post(s)
Liked: 13
Quote:
Originally Posted by hd_newbie View Post

It is not audible until someone proves that it is. Without any evidence, all this is just audiophile boo boo.

Ahh, the debate.

Personally, I haven't trained myself to be a distortion tester. I don't see the upside, and personally I listen to the performance not the recording. Having said that.

As opposed to most 'audiophile boo boo', here we have fundamental reason for a phenomena. There is a mathematical description of the problem. There are measurements confirming the models. We know that the distortion is worse in very quite passages.
We also know that the problem has become obsolete because modern studios record with 24 bit precision, can apply noise shaping, dithers, and makes sure the signal neither clips nor gets lost in the noise. The same is also true in modern players.

Thus, even if quantization distortion is audible, it isn't a problem in commercially produced content. It may or may not be a problem for the home studio, but since the software tools for the home market are relatively robust, it's hard to imagine that it is a problem to any but the most novice garage band recording engineers.
krabapple likes this.
Swampfox is offline  
post #117 of 761 Old 03-12-2012, 12:23 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by Swampfox View Post

Sometimes the two of you argue semantics.

The steps are removed by filtering.
The quantization error is not.
Truncating a digital signal causes truncation error which is quantization error.
Quantization error causes correlated distortion.
Dither essentially decorrelates the quantization distortion.

Thank you so much for posting this. It made me step back and think why there is a disconnect here. The argument shouldn't have been there because dither is an objectivits' best friend as it shows that we don't need deep sample words. Yet here we were with Arny fighting me tooth and nail on it. If I am right, it is not at all about semantics but him fighting a battle that is unrelated and with someone who is on his side already! Let me explain and with it, bring us full circle to the topic at hand.

Let's start at the top. This is how our digital systems work:



As the diagram shows, we take a continuous analog signal, convert it to discreet samples in time, and then transform it back to a continuous analog signal again. The Nyquist–Shannon sampling theorem proves that this transformation works and guarantees to reproduce the original analog waveform. In that sense, it is absolutely true that it is a myth that those discrete steps are a problem. They are not in theory. The two filters at each end of the chain assure that the system works.

Problem is that we have to build real systems. Two things can deviate (at high level) from the ideal assumptions above:

1. That our system bandwidth doesn't exceeds half of our sampling theory. Very sharp filters are needed in the case of CD to accomplish this without harming anything in the audible frequency band. This is unrelated to the argument at hand but was covered earlier by Bob Stuart's paper to have headroom above.

2. The sampling theory assumes that the digital samples have no limitations. As soon as you limit this resolution to say, 16 bits, you now have created a deviation from theory. This has been the core topic of discussion on my part. The "steps" that I discussed were caused by discreet samples values we are talking as opposed to ones with infinite resolution.

Quantization science tells us what happens when we violate #2. We get distortion as we should since we are deviating from ideal representation of those waveform voltages. In time domain, we do get "steps" that are due to our system limitation of not being able to represent analog waveform amplitude with infinite perecision. This phenomenon is unrelated to discrete steps in time. We now have discreet steps in values.

Quantizations steps are errors in sampling and will manifest themselves as such and nothing in the reconstruction filter removes them. If this were not the case, then we could pick any sample depth, such as 1-bit, and still get our signal. Clearly we cannot (assuming no oversampling). Our 1-bit system would have horrendous distortion spikes that are only 9db blow the peak signal itself and would sound awful (again, don't confuse this analogy with oversampling DACs and DSD -- other measures solve this problem there).

If we review what started this argument, you see that I was focused on the quantization issue and not the fact that samples are discreet in time. Arny brought up the issue of noise shaping which only exists because we have quantization error [additional emphasis here]:

Quote:


Quote:
Originally Posted by arnyk View Post

Do you need 24 bits or 20 bits to capture the music on that half-track master? Don't be silly! The noise floor on those great-sounding old tapes is in the 12-13 bit range depending on how you do the digital. With aggressive noise shaping, 11 bits might work.

For others who don’t know what we are talking about, when digital systems use fixed bit depth, you create distortion since analog waveforms do not have such steps. The way to solve this is to add some amount of random data in the form of noise to the signal prior to digitalizing it. That converts distortion into noise which is a good thing. What is not a good thing is that we have now raised the noise floor of our system relative to the source. Noise shaping says instead of adding noise that has equal power in the audible frequency band, we make it so that it has the least amount where the ear is most sensitive and more of it where it is not.

Given the above, the best use of noise shaping is when you have bandwidth above 20 Khz to stuff the noise there. With 44 Khz sampling, your response stops at 22Khz so you are now essentially forced to put the quantization noise in the audible band. This is one of the other reasons it is better to have some headroom in our sampling rate as to give us a place to “park” these dither noise. If you used 88 Khz for example, you now have 50% more bandwidth to push the noise into.

The above is the type of analysis that Bob does in his paper (and then some) to arrive at his recommendations. He puts thought in there combined with excellent grasp of technology and science. Whether you agree with this conclusions or not, it is worth a read. As I have said, I have always used his paper to prove that we don’t need astronomical sample rates and bit depths for transparency. Its use in reverse by Arny is very odd.

Arny took what I said and turned it into me saying that discrete samples in time are the problem which clearly was not in the same planet as what I was discussing. I was talking about quantization error which he started to talk about himself.

So to summarize, steps due to time sampling are not of concern. I never said they were but Arny thought I did and ran with it no matter how many times I brought us back down to discussion of dither/quantization error. On this front, discreet sample *values* can be a problem as the math and measurements and numerous references demonstrate. It is a deviation from sampling theory and therefore, its method of using a reconstruction filter is of no help whatsoever. I showed that with pictures and measurements of the author describing dither and quantization noise.

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #118 of 761 Old 03-12-2012, 12:48 PM
AVS Addicted Member
 
amirm's Avatar
 
Join Date: Jan 2002
Location: Washington State
Posts: 18,083
Mentioned: 3 Post(s)
Tagged: 0 Thread(s)
Quoted: 755 Post(s)
Liked: 434
Quote:
Originally Posted by Ethan Winer View Post

That's certainly true. But the problem is them, not me.

No. The real problem is that they are the customer, and not you . You are the producer. The customer wants to have the same bits you heard in your studio when you produced the product the talent approved. You are insisting on degrading it even though the medium, unlike the CD, no longer requires it.

Quote:


Some people will not be convinced no matter how compelling and logical the argument.

But your argument is not 100% compelling. You are setting a lower bar than Bob Stuart is for example. He also goes through threshold of audibility and psychoacoustics aspects of the topic and arrives at 20 bits of undithered resolution, not 16. Is there some safety margin built into that? Maybe. But that is a good thing.

Quote:


Remember that pathetic exchange at your forum between me and Mike Lavigne when he claimed that digital misses capturing the "essence" of music that vinyl does capture?

What he wants, you can't deliver. He wants analog sound. What others want is what we are discussing. Which is, they want the same master bits as you have in your workstation. That, you can deliver. But you seem to be saying you don't want to do by choice. I don't see the logic in that.

Quote:


His logic and understanding of audio is equal to the unicorns and fairies in my drawing.

There are extremes on both sides of the fence here. For many people 128 Kbps MP3 is the same as CD. That is a unicorn story too and it was actually repeated by Monty in the paper we are discussing here: “It's true enough that a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate.” I saw no one but me object to that. So clearly we pick our battles and then complain about lack of fairness .

Quote:


I'm sure even you understand that digital is far more accurate than vinyl, and misses nothing, even if some people prefer the sound of vinyl. Yet I'm the one you banned rather than him and those who think and throw insults like him.

Hold on a minute . Yes, digital has far better specs than analog. But since we can't assure same recording for both, the fidelity difference can clearly be there. I recently got a Reel to Reel recorder and second gen master tape. I and everyone else we play that combo for immediately gets goosebumps. I have to keep reminding myself that this system has far worse specs than the CD. But boy, does the *content* as I have it in my hand, which in some cases is entirely different mixes, sounds superb.

On getting banned, you never were and are not now. You post just a few weeks ago there so clearly you still have posting privileges. You chose to walk away because we asked you in private to please not take on people all the time in the forum and that focus your contributions in the dedicated area we had given you. You thought that was too restrictive and left. We took no action otherwise.

Quote:


That's a poor analogy, as is the comparison to a chef spitting on food, but I'll humor you anyway. If the distortion is 90+ dB below the music, and saves half the bandwidth and disk space, where's the damage?

The problem is that the customer is always right and he is saying he doesn't care about bandwidth or disk space. He will and is even paying a premium for higher resolution. Yet you keep insisting that it isn't good for him to get the same fidelity you count on in your studio.

Quote:


You do understand that distortion at -90 dB is around 0.003 percent, yes? What speakers do you know of that won't drown that out by two orders of magnitude?

If I record a -30db signal, your distortion is now only 60 db below it. If I turn up the volume to hear it you think I won't hear the distortion? My distortion is now at 0.1%, not .003%. As to speakers distortion, it is *additive* to whatever is in my source. You can't make a case that speaker distortion completely masks quantization noise. They are different animals, created by different things.

We are tasked with building the best systems we can for our customers. We can use science and math as Bob did to show at what level, we can objectively satisfy people. Deviate from that and you are sharing an opinion. It might be an informed and experienced opinion, but it is still an opinion.

Amir
Founder,
To view links or images in signatures your post count must be 0 or greater. You currently have 0 posts.

"Insist on Quality Engineering"

amirm is offline  
post #119 of 761 Old 03-12-2012, 12:57 PM
AVS Addicted Member
 
Chu Gai's Avatar
 
Join Date: Sep 2002
Location: NYC area
Posts: 14,773
Mentioned: 0 Post(s)
Tagged: 0 Thread(s)
Quoted: 160 Post(s)
Liked: 465
I want to hear every last bit on the compressed music that's out there.

"I've found that when you want to know the truth about someone that someone is probably the last person you should ask." - Gregory House
Chu Gai is offline  
post #120 of 761 Old 03-12-2012, 01:25 PM
AVS Addicted Member
 
Ratman's Avatar
 
Join Date: May 2002
Location: Collingswood, N.J.
Posts: 14,455
Mentioned: 1 Post(s)
Tagged: 0 Thread(s)
Quoted: 167 Post(s)
Liked: 285
I also make sure I count and weigh every "bit" in every box to make sure I get what is expected.
LL
Ratman is offline  
Reply Audio theory, Setup and Chat

User Tag List

Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off